Boundary-interior exchanges controlling the labrador sea dynamics

(1)

Boundary-interior exchanges controlling the labrador sea dynamics

Georgiou, S.

DOI

10.4233/uuid:37028134-4e97-46b3-a50b-54ab2cdc25c4

Publication date

2021

Document Version

Final published version

Citation (APA)

Georgiou, S. (2021). Boundary-interior exchanges controlling the labrador sea dynamics.

https://doi.org/10.4233/uuid:37028134-4e97-46b3-a50b-54ab2cdc25c4

Important note

To cite this publication, please use the final published version (if applicable).

Please check the document version above.

Copyright

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy

Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

(2)

B

OUNDARY

-

INTERIOR EXCHANGES CONTROLLING

(3)

(4)

B

OUNDARY

-

INTERIOR EXCHANGES CONTROLLING

THE

L

ABRADOR

S

EA DYNAMICS

Dissertation

for the purpose of obtaining the degree of doctor at Delft University of Technology,

by the authority of Rector Magnificus, prof.dr.ir. T.H.J.J. van der Hagen, chair of the Board for Doctorates,

to be defended publicly on Wednesday 24 March 2021 at 15:00 o’clock

by

Sotiria G

EORGIOU

Master of Science in Environmental Physics, University of Athens, Athens, Greece.

(5)

Composition of the doctoral committee: Rector Magnificus, chairperson

Prof.dr. J.D. Pietrzak, Delft University of Technology, promotor Dr. C.A. Katsman, Delft University of Technology, copromotor Independent members:

Prof.dr.ir. A.J.H.M. Reniers, Delft University of Technology Prof.dr. H. Johnson, University of Oxford, United Kingdom

Prof.dr. P.N. Holliday, National Oceanography Centre, United Kingdom Dr.ir. R. Gelderloos, Johns Hopkins University, United States

Dr. M.F. de Jong, Royal Netherlands Institute for Sea Research Prof.dr.ir. S.G.J. Aarninkhof, Delft University of Technology, reserve member

This research was carried out with the financial support of the Netherlands Organization for Scientific Research (NWO) through the VIDI-Grand 864.13.011.

Printed by: Proefschrift All In One (AIO) Cover by: Christos Georgiou

(6)

Λόγος το καράβι μου πανιά οι λέξεις μου να ταξιδέψω μακριά σ΄ Ιθάκες ονειρεμένες Ανδρέας Χ. Γεωργίου

(7)

(8)

Summary

The Atlantic Meridional Overturning Circulation (AMOC) quantifies the strength of the northward surface transports and southward deeper transports in the Atlantic Ocean. In particular, as warm surface waters flow northward in the Atlantic they gradually cool on they journey releasing heat to the atmosphere. Then, in the subpolar and polar regions these surface waters become dense enough to sink forming cold deep waters, which re-turn southward through the Atlantic Ocean. This oceanic circulation pattern is of high importance for the Earth’s climate since it is the main distributor of heat, biogeochemical tracers and water masses globally. For decades, both the ocean and climate communi-ties have focused on understanding the dynamics that shape the AMOC strength and most importantly its variability. Although, a lot of progress has been made towards this goal, it is still unclear how the AMOC will respond in a changing climate.

Observational and numerical modelling studies are focusing on regions within the Atlantic Ocean that are known to play an important role in shaping the AMOC dynamics. Marginal seas are considered to be such key regions. There the deep waters are formed and then are transported to the global ocean. Although deep water formation regions are located in the interior of the marginal seas it is known that they are not associated with net vertical motions. Instead, several studies suggest that the net downwelling occurs along the perimeter of the marginal seas and that the ocean eddies play an important role in regulating this process. The Labrador Sea, which is located between the west coast of Greenland and the Labrador Peninsula, is known to be one of these key regions in the subpolar North Atlantic. It is characterized by a rich activity of different scale vor-tices (eddies) that act as a bridge between the boundary current system that encircles the basin and its interior. Therefore, a daunting challenge in physical oceanography is to understand how the ocean eddies interact with the large-scale circulation that sets the strength of the AMOC. This thesis aims to provide a better understanding of the role of the interactions between the boundary current and the interior of the Labrador Sea on its dynamics. It employs a combination of an idealized model, a realistic model and obser-vational data to probe the connection between the key physical processes that shape the Labrador Sea dynamics. Furthermore, it examines to what extent changes in the surface heat fluxes can affect these connections and thus the Labrador Sea dynamics.

Over the last decades, there has been significant progress in the numerical modelling of the ocean. Unfortunately, resolving oceanic eddies in global, long simulations is still a challenge for the numerical modelling community. This is due to the fact that a suffi-cient horizontal resolution is required to resolve such processes, which results in com-putationally expensive simulations. Idealized configurations of numerical models allow the use of high horizontal resolution and facilitate the investigation of certain physical processes. In this thesis, an idealized eddy-resolving model is employed to examine the interplay between the ocean eddies, the deep water formation process and the

(11)

welling in the Labrador Sea. The results highlight the necessity of properly resolving the eddy activity in the Labrador Sea in order to represent the downwelling and overturning in the North Atlantic Ocean, and also to predict its response to changing environmental conditions.

There is a limited understanding of how the locally formed water mass leaves the in-terior of the Labrador Sea. An additional objective of this thesis is to investigate the path-ways and the transformation of the water masses in a reference frame that moves along with the flow. This is done by using a Lagrangian particle tracking tool which is initialized by the three-dimensional velocity fields from both idealized and realistic ocean models. In particular, numerical particles are released within the boundary current at the exit of the Labrador Sea and are traced backwards in time. The results presented in this thesis reveal that prior to exiting the domain the water masses follow either a fast, direct route within the boundary current or a slow, indirect route that involves boundary current-interior exchanges. These exchanges are regulated by the presence of mesoscale eddies. Additional analyses of Argo float trajectories confirm the existence of these routes. More-over, it is shown that the water mass transformation along these routes differs. Denser waters are formed in the interior of the basin and follow the indirect route to exit the basin. Instead, lighter waters are formed within or close to the boundary current and fol-low the direct route to leave the basin. This means that boundary-interior exchanges are important for the pathways and the properties of the water masses that exit the Labrador Sea. This underlines the necessity of resolving the mesoscale features required to cap-ture the interior-boundary current exchange in order to correctly represent the export of the newly-formed water masses.

In a changing climate, to accurately make future climate projections, it is necessary to improve our knowledge on how the ocean affects the climate and vice versa. In this thesis, the response of the dynamics of a marginal sea subjected to surface heat loss, such as the Labrador Sea, to changes in the atmospheric forcing is investigated. The results show that complex interactions between the boundary current and interior are established via the eddy activity, and in concert determine the downwelling in the basin as well as the process of deep water formation. Furthermore, the results reveal that in a colder (warmer) regime the strength of the upstream pathways of the water masses exiting the Labrador Sea is affected in a non-linear way and also the transformation of these water masses results in denser (lighter) water masses. The latter, yields a shift of the local overturning maximum in density space towards a higher (lower) density layer.

Taken together, this dissertation emphasizes the importance of the interaction be-tween the boundary current and the interior of the Labrador Sea for the dynamics of the basin. In particular, it highlights the crucial role of the eddies for the transformation and export of deep waters and the associated timescales from the basin to the global ocean. Furthermore, the results presented in this thesis show that the response of the Labrador Sea dynamics to changes in the atmospheric forcing is indirect mainly due to the presence of the rich eddy activity in the basin. Therefore, it is essential to improve our understanding of the boundary current-interior interaction in the marginal seas of the North Atlantic to reduce uncertainty in the estimates of the AMOC and thus, future climate predictions.

(12)

Samenvatting

1

In de Atlantische Oceaan stromen de ondiepe watermassa’s noordwaarts en de die-per gelegen watermassa’s zuidwaarts. Als de warme ondiepe watermassa’s noordwaarts stromen, geven ze warmte af aan de atmosfeer. Hierdoor koelen ze af en daardoor neemt hun dichtheid toe. De dichtheid van deze afgekoelde watermassa’s is in de (sub)polaire regio’s hoog genoeg om de koude diepe watermassa’s te vormen die op diepte terugstro-men in zuidwaartse richting. Deze circulatie is van groot belang voor het klimaat, omdat het de verspreiding van warmte, biochemische tracers en watermassa’s in de oceaan be-paald.

De AMOC (‘Atlantic Meridional Overturning Circulation’) is een graadmeter voor de sterkte van deze grootschalige circulatie. Omdat variaties in de grootte van de AMOC direct gevolgen hebben voor het klimaat, hebben oceaan- en klimaatwetenschappers in de afgelopen decennia geprobeerd om de dynamica en de tijdsvariaties van AMOC te begrijpen. Ondanks dat er op dit gebied al veel vooruitgang is geboekt, is het nog niet bekend hoe de AMOC zal veranderen door klimaatverandering.

Een cruciaal onderdeel van deze circulatie is de connectie tussen de noordwaarts stromende ondiepe en de zuidwaarts stromende diepe watermassa’s. Deze diepe wa-termassa’s ontstaan in verschillende randzeeën van de Atlantische Oceaan. Onderzoek heeft aangetoond dat de eigenschappen van deze diepe watermassa’s ver van de kust bepaald worden, maar dat het verticale transport dicht bij de kust plaatsvindt.

De Labradorzee is zo’n randzee. De Labradorzee ligt in de Noord Atlantische Oce-aan tussen Groenland en het Canadese schiereiland Labrador. Deze zee heeft een breed scala aan oceaanwervels. Omdat de oceaanwervels de sterkte van het verticale trans-port be ¨nvloeden, is het onduidelijk hoe de dynamica in de Labradorzee de sterkte van AMOC be ¨nvloedt. In dit proefschrift wordt gekeken naar de invloed van oceaanwer-vels op de interactie tussen de stromingen langs de kust en de aflandige gebieden in de Labradorzee. Hiervoor worden plaatselijke metingen en modellen met versimpelde en realistische configuraties gebruikt.

De versimpelde configuratie van de Labradorzee wordt gebruikt om te kijken hoe verschillende processen de dynamica be ¨nvloeden. In deze configuratie kan, in tegen-stelling tot lange simulaties in mondiale modellen, een hoge resolutie gebruikt worden. Met deze hoge resolutie kunnen oceaanwervels worden gemodelleerd, en kan dus de wisselwerking tussen de wervels, het ontstaan van diepe watermassa’s en het verticale transport in de Labradorzee onderzocht worden. De resultaten tonen aan dat de oce-aanwervels noodzakelijk zijn voor de totstandkoming van het verticale transport. Deze wervels zijn daarom nodig om accurate voorspellingen te doen van de grootschalige cir-culatie in de Noord-Atlantische Oceaan.

translated in Dutch by Carine van der Boog

(13)

Een combinatie van versimpelde en realistische configuraties wordt gebruikt om te onderzoeken hoe de nieuwgevormde diepe watermassa’s hun ontstaansregio in de La-bradorzee verlaten. Hiervoor wordt een methode gebruikt waarin het referentiekader met de stroming meebeweegt. Dit referentiekader wordt ge ¨mplementeerd met daar-toe ontwikkelde Lagrangiaanse deeltjesvolger die gebruik maakt van driedimensionale snelheidsvelden. Het voordeel hiervan is dat daardoor tegelijkertijd veranderingen in de eigenschappen van deze watermassa’s kan worden onderzocht.

In beide configuraties worden numerieke waterdeeltjes losgelaten in de uitstroom van de Labradorzee en vervolgens terug in de tijd gevolgd. De resultaten laten zien dat, voordat de watermassa’s de Labradorzee uitstromen, deze een directe of indirecte route volgen. De directe route is relatief snel en volgt de krachtige stromingen langs de kust. In de indirecte route is er een uitwisseling met aflandige gebieden. Deze uitwisseling wordt gereguleerd door oceaanwervels. De aanwezigheid van deze routes is bevestigd met behulp van plaatselijke metingen van Argo boeien.

Daarnaast geven de resultaten van de Lagrangiaanse deeltjesvolger en van deze Argo-boeien aan dat eigenschappen van watermassa’s veranderen langs hun pad. De deeltjes die de indirecte route volgen hebben gemiddeld een hogere dichtheid als ze de Labra-dorzee uitstromen. Dit komt omdat deze deeltjes een uitwisseling hebben met de af-landige gebieden waar de diepe watermassa’s met een hoge dichtheid worden gevormd. De deeltjes die de directe route volgen hebben gemiddeld een relatief lage dichtheid. Deze dichtheid komt overeen met de dichtheid van de watermassa’s die ontstaan in de krachtige stroming langs de kust.

Omdat de directe en indirecte route verschillende watermassa’s in de uitstroom re-presenteren, is het belangrijk om uitwisseling tussen het water dichtbij en ver van de kust te begrijpen. Deze uitwisseling bepaald namelijk de verhouding van het aantal deel-tjes die de directe of indirecte route nemen. Deze uitwisseling vindt plaats op de meso-schaal. Daarom is het noodzakelijk om deze processen, waaronder de oceaanwervels, te modelleren. Op die manier kan de uitstroom van watermassa’s uit de Labradorzee correct weergegeven worden.

Het ontstaan van watermassa’s en de sterkte van het verticale transport in de Labra-dorzee wordt be ¨nvloed door de temperatuur van de atmosfeer. In dit proefschrift wordt dit verband verder onderzocht met een model met een versimpelde configuratie. De re-sultaten tonen aan dat een koelere (warmere) atmosfeer resulteert in watermassa’s met een hogere (lagere) dichtheid. Deze dichtheidsverschillen worden onder andere veroor-zaakt doordat de temperatuur van atmosfeer de sterkte van de stromingen reguleert. Deze dichtheidsverschillen veroorzaken, op hun beurt, een verschuiving van het maxi-male verticale transport naar watermassa’s met hogere (lagere) dichtheiden.

Samengevat benadrukt dit proefschrift het belang van de wisselwerking tussen de kuststroming en de aflandige gebieden in de Labradorzee. Er is aangetoond dat de oce-aanwervels een cruciale rol middels hun invloed op het ontstaan, de transformatie en de tijdsvariabiliteit van de watermassa’s. Daarnaast tonen de resultaten in dit proefschrift aan dat atmosfeer de dynamica van de Labradorzee ook indirect be ¨nvloed via de oce-aanwervels. Daarom is het belangrijk om deze wisselwerking nog beter te begrijpen. Op die manier kunnen onzekerheden die geassocieerd worden met voorspellingen van de AMOC worden gereduceerd.

(14)

Chapter 1

Introduction

Give it a whirl1 Sivert Høyem

1.1. The Atlantic Meridional Overturning Circulation

The exploration of the ocean began when people living in coastal areas wanted to use its resources and to explore the unknown. This takes us back thousands of years, to around 7000 B.C, where evidence shows the extensive seafaring and sailing of great distances carried out by civilizations such as the Polynesians, Greeks and Egyptians. Today, know-ing that the ocean controls the Earth’s climate, the exploration of the ocean is motivated by the need to understand its dynamics and to predict the impacts of its variability in a changing climate.

One of the most important systems in the ocean is the Atlantic Meridional Overturn-ing Circulation (or AMOC, Johnson et al.,2019). It quantifies the strength of the north-ward surface transports and southnorth-ward deeper transports in the Atlantic Ocean. Warm surface waters flow northward along vast distances in the Atlantic (Figure 1.1). This cur-rent flow gradually cools on its journey northward releasing heat to the atmosphere. In the subpolar and polar regions the surface waters become dense enough to sink forming cold deep waters, which return southward through the Atlantic Ocean. The northward transport of heat due to the AMOC is responsible for a relatively warm climate in the Northern Hemisphere compared to that of the Southern Hemisphere (Buckley and Mar-shall,2016). Moreover, the AMOC is one of the key processes controlling the uptake and distribution of biogeochemical tracers (i.e., CO2) that are important for the climate

sys-tem (Fontela et al.,2016).

Because of its role for climate, the AMOC is a focus, in both atmospheric and oceano-graphic communities, for understanding present and future climate change. The AMOC is frequently schematized as a conveyor belt (Figure 1.1) implying a laminar and coher-ent ocean transport. However, observing systems that have been deployed throughout the Atlantic Ocean, such as the RAPID array at 26.5oN (Smeed et al.,2018), have revo-lutionized understanding of AMOC variability and have clearly indicated that in reality 1_{https://www.youtube.com/watch?v=4IUR7agj6Mg}

(15)

1

Figure 1.1: Schematic of the Atlantic Meridional Overturning Circulation (AMOC); relatively warm waters (red) flow northwards in the upper layers of the Atlantic Ocean while colder water (blue) returns southwards in deeper layers. Figure is adapted fromMarotzke(2012).

an undisturbed representation of the AMOC as seen inFigure 1.1is not possible. For ex-ample, the existence of the mesoscale ocean variability affects the pathways of both the surface and deep currents that constitute the AMOC and thus, is not compatible with a laminar view (Buckley and Marshall,2016, and references therein). Therefore, to de-termine the strength, structure, and variability of the AMOC a combination of complex physical processes needs to be considered (McCarthy et al.,2020).

Traditionally, the strength of the AMOC is estimated in depth coordinates to empha-size the vertical transport of water masses. It is commonly defined in depth space as the total northward transport in the upper 1000 m or so of the water column (Wunsch and Heimbach,2006;Cunningham et al.,2007). Therefore, AMOC in depth space tends to emphasize sinking (i.e., vertical mass flux). An alternative representation of the AMOC is in density space which emphasizes the transformation of lighter to denser water masses (e.g.,Xu et al.,2016;Brüggemann and Katsman,2019). Deep ocean convection is the process that describes how the surface waters become denser and mix vertically with the deeper layers (Marshall and Schott,1999). There are few regions in the ocean where deep convection occurs. Observations in the marginal seas of the North Atlantic (i.e. the Labrador, the Irminger and the Nordic Seas) reveal convected waters in layers deeper than 1500 m (e.g.,Dickson et al.,1996;Lazier et al.,2002;Pickart et al.,2002;Våge et al.,

2011a;de Jong et al.,2012;Eldevik et al.,2009). Therefore, to emphasize the transforma-tion of lighter to denser water masses a more appropriate estimate of the AMOC strength at high latitudes is in density coordinates (Zhang,2010;Xu et al.,2016;Li et al.,2017; Des-bruyères et al.,2019;Lozier et al.,2019).

(16)

1.2. The role of the Labrador Sea for the AMOC

1

3

The AMOC strength reported in several studies ranges between 16 to 18 Sv (1 Sv = 106 m3s−1,Sarafanov et al.,2012;McCarthy et al.,2015). The timeseries between 2004-2012 at the RAPID array (at 26.5oN) indicated a decline of the AMOC strength of about 0.54 Sv y−1(Smeed et al.,2014). Although, the updated timeseries up to 2018 display an in-creasing tendency of the AMOC starting from 2009, this increase has yet low statistical significance (Moat et al.,2020). Furthermore, over the course of the twenty-first cen-tury, climate models predict a possible weakening of the AMOC’s strength rather than an abrupt transition or shut down (IPCC,2014). Since the AMOC is a key component of the global climate system, a weakening of its strength could lead to abrupt changes in the Earth’s climate (Srokosz et al.,2012). For instance, a weakening AMOC could cause a cooling over the subpolar North Atlantic (e.g.,Drijfhout et al.,2012) and in general the northern hemisphere, an increase of the North Atlantic storms and large changes in pre-cipitation in the tropics (Jackson et al.,2015).

Several studies have suggested a strong connection between the AMOC variability and the process of deep convection in the North Atlantic (Eden and Willebrand,2001;

Rahmstorf et al.,2005;Biastoch et al.,2008). However, recent observational studies have cast doubts on the importance of deep water formation in the Labrador Sea for the AMOC. In particular, measurements along an array across the subpolar North Atlantic (placed north of 52oN) within the Overturning in the Subpolar North Atlantic Program (Lozier et al.,2017) indicated that the contribution of the Labrador Sea to the overturn-ing is very small compared to the eastern part of the subpolar North Atlantic (Lozier et al.,2019;Zou et al.,2020). In this view, the observational study ofChafik and Rossby

(2019) suggested that the transformation of water masses in the Nordic Seas regulates the AMOC strength rather the one in the Labrador Sea. However, the Labrador Sea re-mains a key region in the North Atlantic. It is one of the rare locations where dense waters that contribute to the lower limb of the AMOC are formed (e.g.,Johnson et al.,2019). In addition, the equatorward transport of dense waters that formed elsewhere in the North Atlantic is mainly achieved through the Labrador Sea (Lozier et al.,2013;Bower et al.,

2019). In terms of climate, it is thus important to fully understand the processes that govern the AMOC and its response to projected changes in the formation and spreading of deep water masses.

1.2. The role of the Labrador Sea for the AMOC

The prevailing circulation pattern of the Labrador Sea together with the extreme heat loss that the basin experiences create favorable conditions for the production of dense water masses in the Labrador Sea (Marshall and Schott,1999). In this section, the general circulation of the Labrador Sea is first introduced, followed by the description of the process of deep convection that occurs in the interior of the basin. Furthermore, the observed variability of the properties of deep convection is discussed in this section. Last, this section provides details on the mechanism behind a resulting overturning cell in a marginal sea subject to strong heat loss, such as the Labrador Sea.

(17)

1 1.2.1. General circulation in the Labrador Sea

The Labrador Sea, located between the west coast of Greenland and the Labrador Penin-sula, is a marginal sea of the subpolar North Atlantic. Its circulation can be generally described by a strong boundary current system and a relatively quiescent flow in the interior of the basin (Figure 1.2,Lavender et al.,2000).

East of Greenland the warm, salty Irminger Current (IC) flows alongside the cold, fresh East Greenland Current (EGC). At the southern tip of Greenland the EGC and the IC continue as the West Greenland Current (WGC) flowing on the continental shelf ( Gas-card and Clarke,1983;Fratantoni and Pickart,2007). More accurately, the ECG and IC are thought to continue as the WGC in the Labrador Sea because they have compara-ble velocities (Pickart et al.,2005), but the differences in their thermohaline properties mentioned above still hold. At the northern edge of the Labrador Sea the WGC separates into a part that continues northward to Davis Strait (Curry et al.,2014) and a part that flows westward and joins the cold, fresh outflow from Davis Strait forming the Labrador Current (LC,Lazier and Wright,1993).

Figure 1.2: Map of the subpolar North Atlantic compiled from ETOPO1 data (Amante and Eakins,2009). A schematic of the circulation is shown; EGC and WGC: East and West Greenland Currents, respectively. LC: Labrador Current, IC: Irminger Current, NAC: North Atlantic Current, DSOW: Denmark Strait Overflow Water, ISOW: Iceland Scotland Overflow Water and DWBC: Deep Western Boundary Current.

Below the WGC and the LC, the Deep Western Boundary Current (DWBC) transports both dense waters formed in the subpolar North Atlantic and overflow waters from the Nordic Seas. These overflow waters enter the subpolar North Atlantic by crossing either the Denmark Strait (Denmark Strait Overflow Water, DSOW;Jochumsen et al.,2017) or the Faroe-Bank Channel (Iceland Scotland Overflow Water, ISOW;Hansen et al.,2016).

1.2.2. Deep convection

Every winter, cold outbreaks from the Canadian Arctic create very large heat losses over the ice-free waters of the Labrador Sea (Cuny et al.,2005). The prevailing anti-clockwise

(18)

1.2. The role of the Labrador Sea for the AMOC

1

5

(cyclonic) circulation pattern in the Labrador Sea (Figure 1.2) has a preconditioning ef-fect on the occurrence of deep convections in the basin. Under these circumstances, isopycnals dome upwards in the interior of the basin, which in turn exposes weakly stratified water from the interior to strong surface heat fluxes, reducing the stratifica-tion of the water column through vertical mixing (Marshall and Schott,1999). These heat fluxes together with the prevailing circulation pattern in the basin foster deep con-vection events and with them the formation of the Labrador Sea Water (LSW). This wa-ter mass contains high concentrations of oxygen and other climatically important trace gases from the ocean surface (Rhein et al.,2017). The ventilation of the LSW is the process that describes the transport of these surface signals into the Labrador Sea inte-rior, which subsequently they spread into the intermediate to deep layers of the Atlantic Ocean. This water mass can be ventilated when the wintertime convection reaches suffi-ciently deep (Brandt et al.,2007). Furthermore, LSW is very important since is the upper constituent of the lower branch of the AMOC (e.g.,Rhein et al.,2017).

A diagnostic that is commonly used to study the properties of deep convection is the mixed layer depth (MLD,Figure 1.3). It is usually defined using a density difference criterion: the depth at which the density isρ =0.01 kg m−3higher than the surface value (e.g.,Courtois et al.,2017). In the winters of 2014 and 2015, an extended phase of strong surface heat loss resulted in MLD in 2015 beyond 2100 m (Figure 1.3and Yashayaev and Loder,2017;Piron et al.,2017). Such intensity of convection had not been observed since the beginning of 1990 (Yashayaev and Loder,2017).

Additionally,Figure 1.3indicates that the winter MLD in the Labrador Sea is subject to considerable interranual variability since both the depth and the horizontal expansion of the convection region differ from winter to winter. Estimates of maximum MLD in the Labrador Sea for the period 1993-2009 derived from moorings, ship-based measure-ments and autonomous profiling floats are summarized in the study ofGelderloos et al.

(2013). Their literature review illustrated that the deep convection cycle in the Labrador Sea exhibits considerable interannual variability. This variability is regularly correlated with the strength of the wintertime surface heat loss (Pickart et al.,2002). In the case of subsequent cold winters, the restratification of the basin cannot be fully achieved since the strength of the deep convection is larger, allowing the basin to remember previous winters. However, the response of the deep convection to atmospheric forcing is not straightforward. Observations reveal a rapid warming of intermediate layers after deep convection events that cannot be explained from the surface warming alone (Lilly et al.,

1999), but also requires a contribution from lateral buoyancy fluxes. In particular, several studies suggested that changes in the boundary currents, and thus in the lateral fluxes, can also regulate the variability of the properties of deep convection on interannual time scales (Straneo et al.,2013;Rykova et al.,2015). Therefore, in terms of climate, it is of fundamental importance to understand the mechanisms responsible for the variability of deep convection in the Labrador Sea.

1.2.3. Overturning in the Labrador Sea

The prevailing circulation of the Labrador Sea is considered to be a vital mechanism for its dynamics (Spall and Pickart,2001). The boundary current system in a marginal sea subject to intense heat loss (such as the Labrador Sea), becomes more dense in the

(19)

1

Figure 1.3: Mean March mixed layer depth in the subpolar North Atlantic in (a) 2014, (b) 2015, (c) 2016 and (d) 2017. Black contours denote maximum mixed layer deeper than 1000 m. Data is obtained from the global ocean 1/12o physics analysis and forecast product from the EU Copernicus Marine Services (https://marine.copernicus.eu).

downstream direction (Spall,2004;Straneo,2006a). This densification of the boundary current is achieved by the strong surface heat loss and also by the lateral heat fluxes (Spall and Pickart,2001;Spall,2004;Straneo,2006a;Brüggemann and Katsman,2019,

chapter 2). An important concept of the general circulation in the Labrador Sea rele-vant to climate is the overturning circulation, which results from this densification of the boundary current.

The overturning reflects the amount of water that exits the basin at a different depth than at which it entered due to the densification of the boundary current (Straneo,2006a). The theory behind the mechanism of the overturning in a convective marginal sea re-quires that the boundary current system complies with the thermal wind balance (Spall,

2004). The thermal wind balance, under the geostrophic and hydrostatic balance, is given byequation 1.1in vector form:

f ∂~u ∂z =

g ρo×Ohρ

(1.1)

where f is the Coriolis parameter,~u is the horizontal flow component, g is the gravita-tional acceleration,_Ohis the horizontal gradient operator (i.e.,Ohρ =∂ρ_∂xi +ˆ ∂ρ_∂yj ) andˆ ρ

is the density.

In a marginal sea that is characterized by a buoyant boundary current and a denser interior, the densification of the boundary current results in a reduction of the horizon-tal density difference between the boundary current and the interior in its downstream direction. The thermal wind balance yields that the vertical shear of the velocity of the

(20)

1.3. Boundary current-interior interaction

1

7

boundary current also reduces, sinceequation 1.1simply states that wherever there are horizontal density gradients, the horizontal flow must change with depth. As shown in the idealized studies ofSpall and Pickart(2001),Spall(2004) andStraneo(2006a), an im-mediate consequence of this relation is that the upper part of the boundary current slows down in downstream direction, while the lower part speeds up. This process causes the boundary current to become more barotropic as it encircles the basin (Spall and Pickart,

2001;Straneo,2006a), reflecting a reduction of transport in the upper part of the bound-ary current and an increase of transport in its lower part. That is, an overturning occurs somewhere in the basin.

Observational data support the view of an overturning cell in the Labrador Sea.Pickart and Spall(2007) reported an overturning estimate for the Labrador Sea of 2 Sv using a composite of hydrographic sections in the Labrador Sea over the period of 1990-1997. More recently, the continuous measurements along an array across the Labrador Sea (placed north of 52oN) within the Overturning in the Subpolar North Atlantic Program (OSNAP West,Lozier et al.,2017) provided an overturning estimate in density space of 3.3 Sv from data gathered between August 2014 and April 2016 (Zou et al.,2020). These estimates are in line with the theoretical estimate of the overturning derived in the study ofSpall and Pickart(2001).

The recent idealized numerical study ofBrüggemann and Katsman (2019) raised questions on how these changes in boundary current transport arise. The two-layer con-ceptual model ofStraneo(2006a) suggests that there is a mass transfer from the upper part of the boundary current towards its lower part. This implies that there is a diapy-cnal mass flux, thus water mass transformation, within the stratified boundary current itself. However, idealized and realistic model simulations show that strong net down-ward motions do occur near the boundary (Spall,2004;Katsman et al.,2018;Sayol et al.,

2019). In particular, these studies found that that these downward motions peak well below the mixed layer, where diapycnal processes are known to be small. The highly idealized study ofBrüggemann and Katsman(2019) clearly shows that diapycnal down-welling into the densest isopycnal layers only occurs in the interior and indicates that exchanges between the boundary current and the interior play an important role for the local overturning.

1.3. Boundary current-interior interaction

It becomes apparent from the discussion in the previous sections that the interaction between the boundary current and the interior of the Labrador Sea is an essential com-ponent of the dynamics of the basin. This section elaborates the importance of this inter-action on the key processes that occur in a marginal sea, such as the Labrador Sea that is characterized by a buoyant boundary current and a denser interior and is subjected to strong surface heat loss (Figure 1.4). In particular, this section introduces how this interaction may have an impact on the restratification of the basin after deep convec-tion events due to the heat loss to the atmosphere and also on determining the locaconvec-tion where the downwelling occurs in the basin.

(21)

1

Figure 1.4: A schematic illustrating the key circulation features in a convective marginal sea subjected to sur-face heat loss (adapted fromJohnson et al.,2019): (a) buoyant inflowing boundary current, (b) buoyancy loss to the atmosphere, (c) convective plumes, (d) buoyant eddies shed from the boundary current, (e) denser out-flowing boundary current and (f ) downwelling along the topography.

1.3.1. Restratification after deep convection

The heat that is lost to the atmosphere during winter in the Labrador Sea needs to be replaced, because otherwise the water column will continue to get colder. However, it is known from long-term measurements that this is not happening and that the warming during summer is not sufficient since the solar radiation cannot reach deeper layers. As mentioned insection 1.2.2, the rapid warming of the interior of the Labrador Sea after deep convection events can be only explained by the combination of the surface warming after the wintertime and lateral heat fluxes (Lilly et al.,1999). These lateral heat fluxes can be achieved by the advection of water masses from the buoyant boundary current system into the interior of the basin (e.g.,Prater,2002;Lilly et al.,2003), driven by the abundant eddy activity observed in the Labrador Sea (Figure 1.5).

There is rich literature (e.g.,Prater,2002;Eden and Böning,2002;Lilly et al.,2003;

Katsman et al.,2004;Bracco et al.,2008;Chanut et al.,2008;Gelderloos et al.,2011;

Kawasaki and Hasumi,2014;Rieck et al.,2019) on the role of the eddy activity in the Labrador Sea for the process of restratification. The eddy field in the Labrador Sea is characterized by vortices ranging from small-scale to mesoscale. A great example of the strong eddy activity that is found in the Labrador Sea can be seen inFigure 1.5. The swirls of color visible in the waters delineate eddies which have diameters ranging from a couple of kilometers to a couple of hundred kilometers.

Three types of eddy have been identified in the Labrador Sea (Gelderloos et al.,2011;

Rieck et al.,2019, and references therein): Irminger Rings (IR), convective eddies (CE), and boundary current eddies (BCE). Around the convection region a rim current de-velops and by the process of baroclinic instability sheds the small-scale CE (Jones and Marshall,1997). The formation of the BCE is by the process of baroclinic instability dis-tributed around the basin. In particular, BCE’s are generated due to the density differ-ences between the fresh and cold WGC encircling the Labrador Sea and the interior. The diameter of both the CE and BCE ranges between 10-30 km.

(22)

1.3. Boundary current-interior interaction

1

9

Figure 1.5: Eddy activity in the Labrador Sea as seen in mapped satellite ocean color data. The ocean color is given by the spectral distribution of reflected sunlight and can be used to infer the biological activity that grows in the surface waters. This image was downloaded fromhttps://oceancolor.gsfc.nasa.gov/and created by the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (SNPP) satellite platform on June 23, 2016.

The Irminger Rings (IR) are large eddies with a warm, saline core topped by a colder and relatively fresh near-surface layer (Hátún et al.,2007;de Jong et al.,2014). IRs are shed from the boundary current along the west coast of Greenland where the topogra-phy is steep. It is the combination of instability due to the local bathymetry and the baroclinicity of the boundary current that leads to their formation (Eden and Böning,

2002;Katsman et al.,2004;Bracco et al.,2008). These eddies have a diameter of about 60 km and they carry buoyant water from the boundary current. Although IRs are formed away from the Labrador Sea interior, they are able to transport buoyant water to the con-vected area because they are very energetic (Lilly et al.,2003).

The presence of eddies in the Labrador Sea can restratify the interior of the basin by transporting buoyant water from the boundary current toward the interior (Figure 1.4). The relative importance of the different eddy types for the process of restratification in the Labrador Sea is yet unclear since it depends on their lifespan, vertical extent and how far they can propagate from their formation region towards the interior of the basin. However, it is clear that eddies play a key role in the exchange of heat fluxes between the boundary current and the convective region and thus for the dynamics of the Labrador Sea.

1.3.2. Deep convection versus downwelling

The general idea of the deep convection is that the dense waters are produced in the interior of the marginal seas, where the stratification is weak and the surface waters are exposed to strong heat losses (Marshall and Schott,1999). Deep convection is character-ized by a downward motion within very localcharacter-ized convective plumes which is

(23)

compen-1

sated by upward motion surrounding the plumes (Figure 1.4). Therefore, the net vertical motion over deep convection areas is relatively small (Send and Marshall,1995).

The study ofSpall and Pickart(2001) aimed to have a better understanding of where the waters formed by deep convection sink and what the physical processes are that con-trol this location. They found by using a simple theoretical estimate that a widespread downwelling in a marginal sea at high latitudes would have to be balanced by an unre-alistically large horizontal circulation. Instead, they suggested that the downwelling of water can occur in the boundary current system along the perimeter of the marginal seas where the dynamical constraints do not hold (Figure 1.4). This implies that the connec-tion between the convective region and the surrounding circulaconnec-tion must also be taken into account.

A few years later,Spall(2004) andStraneo(2006a), using idealized models, examined the interplay of the convective region with the surrounding circulation in a marginal sea. The aim of these studies was to understand how the dense water is removed from the convection region and how the annual mean heat losses to the atmosphere are com-pensated.Spall(2004) found that significant downwelling can occur at the topographic boundaries.Straneo(2006a) showed also that downwelling occurs in the boundary cur-rent as a result of mass conservation, while remaining in geostrophic balance. Both stud-ies found that a cyclonic boundary current balances the heat loss in the interior by ad-vecting buoyant water towards the interior, where the convection occurs, and draining the dense water out. This heat transfer between the boundary current and the interior is assumed to be mainly driven by eddies that are generated from instabilities of the boundary current.

The fact that there is a strong signal of downwelling motion in a narrow region close to the boundary has been shown in numerous idealized regional model studies (e.g.,

Pedlosky and Spall,2005;Spall,2004,2010,2011) and in laboratory experiments (Cenedese,

2012). In addition, most recent global ocean model studies (Katsman et al.,2018; Wald-man et al.,2018;Sayol et al.,2019) have also shown that net downwelling occurs in a narrow region near the topographic boundaries. In particular, Katsman et al.(2018) demonstrated that this downward motion in the Labrador Sea is not uniformly spread along the boundaries but has a stronger signal in the areas where the mesoscale eddy activity is enhanced.

Furthermore,Brüggemann and Katsman(2019) building on the existing studies re-garding the downwelling in depth space (i.e., mass transport from a shallower to greater depth layer;Spall,2004,2010) and in density space (i.e., mass transport from a lighter to denser isopycnal layer;Brandt et al.,2007;Xu et al.,2018) emphasized the important distinction between these two processes. Note that the downwelling in density space is associated with water mass transformation since there is a transport from an isopycnal layer to another and it is also referred as diapycnal downwelling.Brüggemann and Kats-man(2019) concluded that while there is a strong near-boundary downwelling signal in depth space, there is no substantial downwelling in density space in this region. On the contrary, they showed that the diapycnal downwelling mainly occurs in the interior of the marginal sea and along the boundary current but at a different location where there is strong downwelling in depth space.

(24)

1.4. Research objectives

1

11

1.4. Research objectives

As outlined in the previous sections, climate models predict that the AMOC will slow over the next centuries. What hinders the complete understanding of the processes that gov-ern the AMOC and its response to projected climate changes, is a better understanding of the various physical processes that compose the North Atlantic circulation. A number of idealized studies (Spall,2004;Katsman et al.,2004;Straneo,2006a;Gelderloos et al.,

2011;Brüggemann and Katsman,2019) suggest that the presence of mesoscale eddies is crucial in determining the deep convection, downwelling and the characteristics of the circulation in marginal seas subjected to strong surface heat loss. So far, the climate models that are used to make the AMOC projections and show a significant decline over the 21stcentury, have a resolution on the order of 1o(∼100 km) and hence are not able to appropriately resolve the mesoscale eddies. In the absence of these eddies, climate mod-els may misrepresent the deep convection properties and the surface fluxes, adding an uncertainty on the AMOC projections using those models (Randall et al.,2007). In addi-tion, the quantitative impacts of the eddies on the local downwelling and their interplay with the interior is still an open question. Although idealized eddy-resolving models pro-vide helpful qualitative insights on this interplay, to address this question eddy resolving realistic models are more appropriate.

A question that arises from the previous section is whether the strong downwelling signal in the Labrador Sea is associated with dense water masses that either formed in the boundary current itself in this region or elsewhere in the basin. Although, the ideal-ized study ofBrüggemann and Katsman(2019) clearly indicated that the location where these waters are formed differs from the location where the strong downwelling occurs (section 1.3.2), it is yet unclear how these locations are connected. To address this ques-tion in detail a better understanding of the pathways that the LSW follows from its for-mation region to the global ocean is needed. While there are several studies that suggest possible export routes of the LSW (e.g.,Talley and McCartney,1982;Lavender et al.,2000;

Straneo et al.,2003;Brandt et al.,2007;Palter et al.,2008;Bower et al.,2009;Rhein et al.,

2015;Feucher et al.,2019), the various pathways and the associated timescales are yet to be clarified.

Monitoring the AMOC at a specific latitude as the RAPID array (at 26oN,Smeed et al.,

2018) has resulted in reliable estimates for its current strength at this latitude. How-ever, at this latitude, it is difficult to investigate the connection between the processes involved in the deep water formation and the AMOC variability. Observations of the AMOC in the subpolar North Atlantic along the OSNAP array (north of 52 oN,Lozier et al.,2017) aim to improve our understanding on buoyancy-forced AMOC variations. Recent studies based on observations obtained from the OSNAP array suggested a mini-mal contribution of the Labrador Sea to the AMOC strength (Lozier et al.,2019;Zou et al.,

2020). Yet it is unclear what the time frame of these observations should be to accurately make assumptions on possible connections. Furthermore, as discussed in the previous sections, boundary current-interior exchanges in the Labrador Sea might obscure the connection between the deep water formation and AMOC variability in short-term ob-servations such as the OSNAP array.

(25)

1

In this thesis, a combination of numerical modelling techniques and observational data is used to provide new insights on the relationship between the formation, export and associated timescales of deep water masses, and the variability therein. In partic-ular, this thesis aims for a better understanding on the role of the boundary current-interior exchange on the processes that are crucial to determine the AMOC strength and its variability. Hence, this thesis is centered around the following question:

What role does the interaction between the boundary current and the interior of the Labrador Sea play for its dynamics?

1.5. Thesis outline

This section provides an outline of the remaining chapters of this thesis (chapter 2

-chapter 6). Each chapter deals with the research question that formulated in the pre-vious section.

Chapter 2explores the influence of the eddies on deep convection properties in the basin. In addition,chapter 2identifies the regions where the net downwelling of waters occurs and to advance our understanding of the physics that govern these processes. The analysis performed inchapter 2is based on a high-resolution idealized numeri-cal model for the Labrador Sea. Idealized regional models have been identified as ap-propriate tools to represent the processes that occur in the marginal seas in a realistic way (Spall,2004,2010,2011;Katsman et al.,2004;Straneo,2006a;Gelderloos et al.,2011;

Våge et al.,2011b). Moreover,chapter 2deals with the response of deep convection and downwelling to changes in the surface heat fluxes. Although the main method followed inchapter 2uses an Eulerian framework, an attempt to track where the newly-formed dense water goes after leaving its formation region is also presented inchapter 2. To do so, a passive tracer is initialized within the area of deep convection and the areas with high concentration of this tracer are identified.

Chapter 2and also the study ofBrüggemann and Katsman(2019) indicate that the convected water is transported from the interior of the basin towards the boundary and in particular the regions where the eddy activity is enhanced. Strongly motivated by this result, a Lagrangian analysis is performed inchapter 3. This analysis investigates the origin and transformation of different water masses found within the boundary current at the exit of the Labrador Sea and assesses the role of the prominent eddy field for ex-porting these water masses. The Lagrangian analysis followed inchapter 3is based on analyzing the trajectories of numerical particles that are advected using the output from the idealized model applied inchapter 2. In particular, the idealized model study in a La-grangian framework that is followed inchapter 3provides helpful insights on the nature of the interior-boundary current exchange and its role for the LSW export.

The approach and analyses followed inchapter 4, strongly build on the qualitative in-sights gained inchapter 3.Chapter 4delves deeper into the possible export pathways of the water masses exiting the Labrador Sea using both observational data and the output of a realistic global ocean model. The trajectories of available Argo floats that drift within the Labrador Sea are analyzed to identify pathways and associated timescales of deeper water masses within the Labrador Sea from an observational point of view. The use of

(26)

1.5. Thesis outline

1

13

output from a realistic simulation introduces among others more realistic atmospheric forcing and bathymetry, the effects of salinity, potential connections between the sub-basins of the subpolar North Atlantic and the presence of overflow waters. These aspects were not included in the idealized model used inchapter 2andchapter 3. Thus, the methodology followed in this chapter allows a more realistic and quantitative character-ization of the transports and water mass transformation along different pathways.

Chapter 5contains a sensitivity study of the pathways that water masses follow prior to exiting the Labrador Sea with respect to variations in the surface heat loss. The re-sponse of the deep convection to changes in the surface heat fluxes that has been an-alyzed inchapter 2already suggested a rather complicated connection between dense water formation, eddies and surface forcing. Chapter 5investigates how the export of the convected water and the associated timescales are affected by the changes in the surface heat flux. The methodology followed in this chapter combines the output of the idealized sensitivity simulations performed inchapter 2with a Lagrangian tracking tool. Knowledge on the response of the pathways of the convected water in a changing climate is important to direct further observational and modelling efforts.

Chapter 6presents the conclusions of this thesis, reflecting about the role of the boundary current-interior exchange on the dynamics of the Labrador Sea along with potential directions for future research.

(27)

(28)

Chapter 2

On the interplay between downwelling, deep convection

and mesoscale eddies in the Labrador Sea

In this study, an idealized eddy-resolving model is employed to examine the interplay be-tween the downwelling, ocean convection and mesoscale eddies in the Labrador Sea and the spreading of dense water masses. The model output demonstrates a good agreement with observations with regard to the eddy field and convection characteristics. It also dis-plays a basin mean net downwelling of 3.0 Sv. Our analysis confirms that the downwelling occurs near the west Greenland coast and that the eddies spawned from the boundary cur-rent play a major role in controlling the dynamics of the downwelling. The magnitude of the downwelling is positively correlated to the magnitude of the applied surface heat loss. However, we argue that this connection is indirect: the heat fluxes affect the convection properties as well as the eddy field, while the latter governs the Eulerian downwelling. With a passive tracer analysis we show that dense water is transported from the interior towards the boundary, predominantly towards the Labrador coast in shallow layers and towards the Greenland coast in deeper layers. The latter transport is steered by the presence of the eddy field. The outcome that the characteristics of the downwelling in a marginal sea like the Labrador Sea depend crucially on the properties of the eddy field emphasizes that it is essential to resolve the eddies to properly represent the downwelling and overturn-ing in the North Atlantic Ocean, and its response to changoverturn-ing environmental conditions.

This chapter has been published in Ocean Modelling 135. (2019) 56-70.

(29)

2 2.1. Introduction

The Atlantic Meridional Overturning Circulation (AMOC) quantifies the zonally inte-grated meridional volume transport of water masses in the Atlantic Ocean. A prominent feature of the AMOC is an overturning cell where roughly 18 Sv (1 Sv = 106m3s−1_,

Cun-ningham et al. 2007;Kanzow et al. 2007;Johns et al. 2011) of water flows northward above 1000 m, accompanied by a southward return flow at depth. As the surface waters flow northward through the Atlantic Ocean, they become dense enough to sink before they return southward at depth.

This lower limb of the AMOC contains water masses that can be traced back to spe-cific deep ocean convection sites (Marshall and Schott,1999). There are few regions in the world oceans where deep convection occurs, and numerous studies have revealed that the most important ones are in the marginal seas of the North Atlantic (Dickson et al.,1996;Lazier et al.,2002;Pickart et al.,2002;Eldevik et al.,2009;Våge et al.,2011a;

de Jong et al.,2012;de Jong and de Steur,2016b;de Jong et al.,2018).

Through the process of deep convection, dense waters are produced in the interior of the marginal seas, where the stratification is weak and the surface waters are exposed to strong heat losses (Marshall and Schott,1999). While convection involves strong vertical transports of heat and salt, the interior of these marginal seas is known for a negligible amount of net downwelling. In particular, by applying the thermodynamic balance and vorticity balance to an idealized setting,Spall and Pickart(2001) pointed out that in a geostrophic regime, widespread downwelling in the interior of a marginal sea at high lat-itudes is unlikely, as it would have to be balanced by an unrealistically strong horizontal circulation. Instead, substantial downwelling of waters may occur along the perimeter of the marginal seas where the geostrophic dynamical constraints do not hold.

Using an idealized model,Spall(2004) demonstrated that significant downwelling indeed only occurs at the topographic slopes of a marginal sea subject to buoyancy loss. This downward motion yields an ageostrophic vorticity balance in which the vertical stretching term and lateral diffusion term near the boundary dominate (Spall,2010).

Straneo(2006a) considered the downwelling near the boundary from a different per-spective, by developing an analytical two-layer model. In this study, a convective basin is represented by two regions; the interior, where dense water formation occurs due to surface buoyancy loss, and a buoyant boundary current that flows around the perime-ter of the marginal sea. It is assumed that instabilities provide the laperime-teral advection of buoyancy from the cyclonic boundary current towards the interior required to balance the atmospheric buoyancy loss over the interior. This alongstream buoyancy loss of the boundary current reduces the density difference between the boundary current and the interior along the perimeter of the marginal sea. As a consequence, the thermal wind shear of the boundary current decreases in downstream direction, and continuity then demands the water to downwell at the coast (see alsoKatsman et al.(2018) and refer-ences therein).

Spall and Pickart(2001) argue that the magnitude of the buoyancy loss of the bound-ary current determines the amount of downwelling that occurs near the boundbound-ary. While the surface buoyancy loss contributes to this buoyancy loss, it is assumed to be mainly driven by eddies generated by instabilities of the boundary current (Spall,2004;Straneo,

(30)

2.1. Introduction

2

17

2006a).

Eddies shed from the boundary current also play an important role for the cycle of ocean convection and restratification. Deep convection occurs during wintertime in the southwest Labrador Sea (Clarke and Gascard,1983;Lavender et al.,2000;Pickart et al.,2002;Våge et al.,2008). The dense water that is formed during the convection events, Labrador Sea Water (LSW), strongly contributes to the structure of the North At-lantic Deep Water, which in turn is a crucial component of the AMOC (Lazier et al.,2002;

Yashayaev et al.,2007;Pickart and Spall,2007;Lozier,2012). Several studies show that the thermohaline characteristics of LSW are influenced not only by external parameters like the surface heat fluxes, but also by the baroclinic structure of the boundary current that enters the Labrador Sea (Spall,2004;Straneo,2006b), known as the West Greenland Current (WGC), and its interannual variability (Rykova et al.,2015).

In the Labrador Sea heat is carried from the WGC into the interior by Irminger Rings (IRs): large mesoscale eddies that are formed off the west coast of Greenland in a region characterized by a steep topographic slope (Lilly et al.,2003;Katsman et al.,2004;Bracco et al.,2008;Gelderloos et al.,2011). It has been recognised that the IRs strongly con-tribute to compensating the annual mean heat loss to the atmosphere that occurs in the Labrador Sea (Katsman et al.,2004;Hátún et al.,2007;Kawasaki and Hasumi,2014).

From the above, it is clear that eddies are of immense significance for the down-welling as well as for the convection and the heat budget in the Labrador Sea. The dy-namics of the downwelling and how it is related to the observed export of dense water masses is a topic of ongoing research, as the quantitative effects of the interplay between downwelling, eddies and convection are far from clear. For example, in a basin subject to buoyancy loss, one expects that an increase of the heat loss will result in denser and most likely deeper mixed layers. At first glance, this will increase the horizontal density gradients within the basin, strengthen the baroclinic instability of the boundary current and hence intensify the eddy field and the strength of the downwelling. This suggests a positive feedback of the increased eddy fluxes on the downwelling. However, the en-hanced efficiency of eddies to restratify the interior after convection may provide a neg-ative feedback on the convection and it is not clear a priori what the net effect will be.

Moreover, observations show that convected waters that originate from the Labrador Sea contribute to the lower limb of the AMOC (Rhein et al.,2002;Bower et al.,2009). This suggests that there has to be a connection between the convective regions (where these dense waters are formed) and the surrounding circulation near the boundary (where waters can sink) that has not been fully explored. Eddies provide a possible natural con-nection between these two regions.

The aim of this chapter is to assess the quantitative impacts of the eddy field on the downwelling in the Labrador Sea and its interaction with deep convection. We seek to gain more insight in the dynamics that control the downwelling in a convective marginal sea and its response to changing forcing conditions. Towards this goal, we use a highly idealized configuration of a high-resolution regional model in order to isolate specific processes and connect the outcomes with theory. In particular, we diagnose how the eddy field influences the downwelling by exchanging heat between a warm boundary current and a cold interior basin subject to convection. We compare our results to previ-ous theories of downwelling dynamics. In addition, we use a passive tracer study to shed

(31)

2

light into the pathways of the dense water masses and especially focus on the role of the eddies in determining these pathways. Finally, by using two sensitivity studies reflecting a milder and colder winter climate state, we test the sensitivity of the downwelling and the export of dense waters with regard to varying surface forcing.

The chapter is organized as follows: the model setup and the simulations performed are described insection 2.2. The representation of deep convection and the character-istics of the downwelling are described insection 2.3. The response of the deep con-vection and the time mean downwelling to changes in the surface forcing is presented insection 2.4, followed by a discussion insection 2.5. The conclusions of this work are presented insection 2.6.

2.2. Model setup

2.2.1. Model domain and parameters

The numerical simulations performed in this study are carried out using the MIT general circulation model (Marshall et al.,1997) in an idealized configuration for the Labrador Sea. MITgcm solves the hydrostatic primitive equations of motion on a fixed Cartesian, staggered C-grid in the horizontal. The configuration of the model is an improved ver-sion of the one used in the idealized studies ofKatsman et al.(2004) andGelderloos et al.

(2011), which now incorporates seasonal variations of both the surface forcing and the boundary current and enhanced vertical resolution.

The model domain is 1575 km in the meridional direction and 1215 km in the zonal direction. It has a horizontal resolution of 3.75 km in x and y direction (Figure 2.1a), which is below the internal Rossby radius of deformation for the first baroclinic mode in the Labrador Sea (∼7.5 km,Gascard and Clarke 1983). The model has 40 levels in the vertical with a resolution of 20 m in the upper layers up to 200 m near the bottom. The maximum depth is 3000 m and a continental slope is present along the northern and western boundaries (Figure 2.1a). FollowingKatsman et al.(2004) andGelderloos et al.

(2011), we apply a narrowing of the topography to mimic the observed steepening of the slope along the west coast of Greenland, which is crucial for the shedding of the IRs from the boundary current (Figure 2.1a,Bracco et al. 2008). The continental shelves are not included. There are two open boundaries (each roughly 100 km wide), one in the east and one in the southwest, where the prescribed boundary current enters and exits the domain. All the other boundaries are closed (Figure 2.1a).

Subgrid-scale mixing is parameterized using Laplacian viscosity and diffusivity in the vertical direction and biharmonic viscosity and diffusivity in the horizontal direction. The horizontal and vertical eddy viscosity are Ah= 0.25 ×109m4s−1and Av= 1.0 ×10−5

m2s−1respectively, while the horizontal diffusion coefficient is Kh= 0.125 × 109m4s−1.

The vertical diffusion coefficient is described by a horizontally constant profile which decays exponentially with depth as Kv(z) = Kb+ K0× e(−z/zb), where Kb= 10−5m2s−1, K0

=10−3m2s−1and zb = 100 m. Temperature is advected with a quasi-second order

Adams-Bashforth scheme. In case of statically unstable conditions, convection is parameterized through enhanced vertical diffusivity (Kv = 10 m2s−1). A linear bottom drag with

(32)

2.2. Model setup

2

19

(a) (b) (c)

IR

Figure 2.1: (a) Snapshot of the sea surface temperature (SST) for the reference simulation (referred to in the text as REF). Black contours outline the bathymetry, the contour interval is 500 m starting from the isobath of 500 m. The grey arrows represent the inflow/outflow, where the boundaries are open (xinflow=1215 km, yinflow=

978.75-1083.75 km and xoutflow= 611.25-708.75 km, youtflow= 3.75 km) (b) Section across the inflow region

(at x = 1215 km) showing the annual mean temperature (Tin, in◦C) and meridional velocity (Uin, in m s−1,

black contours). (c) Zonal section of an example Irminger Ring in the REF simulation by means of temperature (shading, in◦C) and meridional velocity (black contours, in m s−1). This Irminger Ring is visible in the SST snapshot in the basin interior (x=443-525 km and y=1012.5 km) in (a).

FollowingKatsman et al. (2004), the model is initialized with a spatially uniform stratification,ρref(z), representative of the stratification in the western Labrador Sea in

late summer along the WOCE AR7W section. Only temperature variations are consid-ered in the model, so this stratification is represented by a vertical gradient in tem-perature, Tref(z), calculated from ρref(z) using a linear equation of state: ρref(z) =ρ0

[1 − α (T − Tr e f(z))], whereρ0= 1028 kg m−3 andα the thermal expansion coefficient

(α = 1.7 × 10−4 ◦C−1).

The effects of salinity are not incorporated in the model. In reality, salinity does affect the properties of deep convection in the Labrador Sea, as the IRs shed from the bound-ary current carry cold, fresh shelf waters at their core (e.g.,Lilly and Rhines,2002;de Jong et al.,2016). As shown in for exampleStraneo(2006b) andGelderloos et al.(2012), the contribution of this lateral fresh water flux to the buoyancy of the Labrador Sea in-terior impacts the convection depth, and large salinity anomalies may in fact be partly responsible for observed episodes when deep convection shut down (Belkin et al.,1998;

Dickson et al.,1988). However, since we focus here on the underlying dynamics that control the downwelling and its response to changing forcing conditions, the effects of salinity are omitted in the model for simplicity.

2.2.2. Model forcing

At the eastern open boundary, an inflow representing the WGC is specified by a merid-ional temperature field Tin(y, z) and a westward flow Uin(y, z) in geostrophic balance with

this prescribed temperature (Katsman et al.,2004). Although the WGC consist of cold, fresh Arctic-origin waters and warm, salty waters from the Irminger Current ( Fratan-toni and Pickart,2007) we only incorporate in the model density variations associated

(33)

2

with the latter part. The cool, fresh surface waters are omitted, since they are found on the continental shelf, which is not included in our idealized bathymetry (Figure 2.1a). The time-mean structure of this warm boundary current is shown inFigure 2.1b. The boundary current follows the topography and flows cyclonically around the basin. The seasonal variability of the WGC seen in observations (Kulan and Myers,2009;Rykova et al.,2015) is represented in the model by adding a sinusoidal seasonally varying term to the inflow conditions based on these observations (∆Umax= 0.4 cm s−1that peaks in

September and attains its minimum in March). At the southern open boundary an Or-lanski radiation condition (Orlanski,1976) for momentum and tracers is applied.

At the surface, only a temporally and spatially varying surface heat flux is applied, which is an idealized version of the climatology of WHOI OAFlux project (Yu et al.,2008). The strongest heat loss occurs on the northwestern side of the basin (Figure 2.2), and its amplitude decays linearly away from this heat loss maximum (white marker in Fig-ure 2.2b). The net annual heat loss over the entire model domain of the reference sim-ulation (hereinafter REF) is -18 W m−2. The time dependence of the amplitude of the surface heat fluxes (Figure 2.2c) is also based on the observations, ranging from -320 W m−2(January) to 140 W m−2(July) at the location of the heat loss maximum.

(a)

(c)

(b)

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

-400 -300 -200 -100 0 100 200 Q (W m -2) COLD REF WARM

Figure 2.2: (a) 1983-2009 mean of heat flux from the climatology of WHOI OAFlux project (Yu et al.,2008). (b) Annual mean surface heat flux applied to the model. Q< 0 means cooling of the sea surface, Q in W m−2. (c) Seasonal cycle of the amplitude of the heat flux at the location where the amplitude is maximum (white marker in b, solid lines) and the mean over the basin (dashed lines) for the three different simulations (black: REF, red: WARM and blue: COLD).

Boundary-interior exchanges controlling the labrador sea dynamics